Nuance Labs, a 24-person Seattle startup building AI that reads human expression in real time, closed a $50 million Series A round. The company wants to make AI avatars stop staring blankly at people.

Founded in early 2025 by former Apple researchers Fangchang Ma, Edward Zhang and Karren Yang, Nuance Labs is developing a single AI model that reads a person's tone, gaze, gestures and timing during a live conversation and responds with its own facial and vocal expressions.

GeekWire first reported the funding on Tuesday, Sept. 15.

Lightspeed Venture Partners, a returning backer, led the round. Returning investors Accel and South Park Commons also participated, alongside new investors Nvidia and Define Ventures.

The raise brings Nuance Labs' total funding to $60 million, following a $10 million seed round led by Accel in 2025.

Most AI avatar systems chain together separate tools for transcription, text generation and voice output.

Ma, the company's CEO, told Business Insider that approach produces awkward results. "If you use existing AI avatars, when they're listening, they're not reacting, or they're just doing random things," he said.

Nuance built a single "full-duplex" model instead. Audio and video go in; audio and video come out. According to the company's announcement, the model reads a speaker's words, tone, eye contact, gestures and timing, then generates its own facial and vocal responses on the fly.

Ma and Zhang met at Apple's Seattle engineering office, where they worked on digital personas for the Vision Pro headset. Ma holds a PhD from MIT in robotics and machine learning. Zhang earned his computer graphics PhD from the University of Washington.

That UW connection matters for hiring. Ma and Zhang have said they want Nuance "to be the premier research lab in Seattle," choosing to build here rather than in Silicon Valley. The company plans to use the new capital to hire researchers and engineers across modeling, data, evaluation and real-time serving.

Ma said the training data for Nuance's models cannot be scraped from the internet. The company collects data in-house or from data partners, according to Business Insider.

The raise lands during a shift in Seattle's venture landscape. Seattle-area startups raised $2.7 billion in venture funding through the first half of 2026, down about 40% from $4.5 billion in the same period a year earlier, according to PitchBook-NVCA data reported by GeekWire.

AI companies accounted for 86% of all U.S. venture dollars during that stretch.

Nuance targets sales, customer service, coaching and education. The company plans to open a public research preview of its model later in 2026. No specific date has been announced.